Download
statistical machine translation part iii phrase based smt decoding n.
Skip this Video
Loading SlideShow in 5 Seconds..
Statistical Machine Translation Part III – Phrase- based SMT / Decoding PowerPoint Presentation
Download Presentation
Statistical Machine Translation Part III – Phrase- based SMT / Decoding

Statistical Machine Translation Part III – Phrase- based SMT / Decoding

159 Views Download Presentation
Download Presentation

Statistical Machine Translation Part III – Phrase- based SMT / Decoding

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. Statistical Machine TranslationPart III – Phrase-based SMT / Decoding Alex Fraser Institute for Natural Language Processing University of Stuttgart 2008.07.23 EMA Summer School

  2. Outline • Phrase-basedtranslation • Log-linear model • Tuning log-linear model • Decoding

  3. Slide fromKoehn 2008

  4. Slide fromKoehn 2008

  5. Language Model • Usually a trigramlanguage model isusedfor p(e) • P(the man wenthome) = p(the | START) p(man | START the) p(went | the man) p(home | man went) • Language modelswork well forcomparingthegrammaticalityofstringsofthesame length • However, whencomparingshortstringswithlongstringstheyfavorshortstrings • Forthisreason, a veryimportantcomponentofthelanguage model isthelengthbonus • Thisis a constant > 1 multipliedforeach English word in thehypothesis

  6. d ModifiedfromKoehn 2008

  7. Slide fromKoehn 2008

  8. Slide fromKoehn 2008

  9. Slide fromKoehn 2008

  10. Slide fromKoehn 2008

  11. Slide fromKoehn 2008

  12. Slide fromKoehn 2008

  13. Slide fromKoehn 2008

  14. Slide fromKoehn 2008

  15. Slide fromKoehn 2008

  16. Slide fromKoehn 2008

  17. Slide fromKoehn 2008

  18. Outline • Phrase-basedtranslation • Log-linear model • Tuning log-linear model • Decoding

  19. Slide fromKoehn 2008

  20. Slide fromKoehn 2008

  21. Slide fromKoehn 2008

  22. Slide fromKoehn 2008

  23. Slide fromKoehn 2008

  24. Slide fromKoehn 2008

  25. Slide fromKoehn 2008

  26. Slide fromKoehn 2008

  27. Outline • Phrase-basedtranslation model • Log-linear model • Tuning log-linear model automatically • Decoding

  28. Outline • Phrase-basedtranslation model • Log-linear model • Tuning log-linear model automatically • Decoding • Basic phrase-baseddecoding • Dealingwithcomplexity • Recombination • Pruning • Future costestimation • Decoding output

  29. Slide fromKoehn 2008

  30. Slide fromKoehn 2008

  31. Slide fromKoehn 2008

  32. Slide fromKoehn 2008

  33. Slide fromKoehn 2008

  34. Slide fromKoehn 2008

  35. Slide fromKoehn 2008

  36. Slide fromKoehn 2008

  37. Slide fromKoehn 2008

  38. Slide fromKoehn 2008

  39. Slide fromKoehn 2008

  40. Slide fromKoehn 2008

  41. Slide fromKoehn 2008

  42. Slide fromKoehn 2008

  43. Slide fromKoehn 2008

  44. Slide fromKoehn 2008

  45. Slide fromKoehn 2008

  46. Slide fromKoehn 2008

  47. Slide fromKoehn 2008

  48. Slide fromKoehn 2008

  49. Slide fromKoehn 2008

  50. Slide fromKoehn 2008